1 INTRODUCTION
Traditionally, most freshwater mollusks inventories were based on shells records (ex. Duncan, 2008). The durability of shells may be an advantage for species detection but may lead to potential biases. Although shell surveys are less time consuming, they often overestimate population sizes, given that there is always more dead than alive specimens. More importantly, they can also be misleading regarding species occurrences. We have many examples of species being extirpated from a site where shells can still be found for many years. Furthermore, the traditional classification of freshwater mollusks is based on plastic shell traits that make identification difficult without taxonomical expertise. This turns surveys based on shells only a difficult task for some species and a source of numerous species misidentifications. This is especially true for large and conspicuous species such as the Unionoid bivalves. For smaller shells, like the fingernail clams, another important bias in rivers, although poorly documented, is shell drifting, where shells can be collected far away from their source population.
Therefore, most freshwater mollusks atlases based on these classical surveys are likely biased. In these publications, misidentified specimens, ancient data, and shell-only data are pooled together generally leading to enlarged species distributions and improved freshwater mollusks conservation status. Our results based on a large-scale environmental DNA (eDNA) investigation shine a light on these biases and contribute to more reliable species occurrences and therefore to more accurate conservation status assessments.
DNA barcodes provide more reliable species identifications than shell morphology (Prié, Puilandre, & Bouchet, 2012; Prié & Puillandre, 2014; Riccardi et al., 2019). DNA analyses are sometimes used to confirm protected species determination, e.g. for the Thick-shelled River MusselUnio crassus Philipsson, 1788, a protected species in European countries, which is sometimes difficult to identify but may have important economic impacts, if found in an area planned for project development. However, several difficulties hamper the use of DNA barcodes for species identification, among which lab availability, price, permits to sample protected species, as sampling tissue for DNA can be hazardous for mollusks, and the need to find and collect live specimens, which can be a difficult and time-consuming task. Freshwater mollusks are hard to survey, sometimes difficult to detect, especially for small species or species that spend most of their life buried. Freshwater bivalves surveys often involve scuba diving, which means having specialized biologists, costly equipment, and also increased associated risks (e.g. navigation, drifting objects, deadwood, fishing lines, etc…) which have to be accounted for, especially in a professional context.
The development of non-invasive genetic monitoring methods with the extraction of environmental DNA (eDNA) from indirect sources, such as fur, scats, or soil and water samples, brought important advances in the remote detection and monitoring of species that are rare and/or difficult to capture or sample (Taberlet, Coissac, Hajibabaei, & Rieseberg, 2012). eDNA metabarcoding analyses are therefore a promising way to overcome the difficulties regarding freshwater mollusks’ surveys and sampling. They rely on a barcoding approach, which secures determinations. Samples can be easily and safely collected in the field (contrarily to scuba diving or river prospecting), do not require particular skills, and provide data about the extant living population, as eDNA cannot be detected for long in freshwater (Dejean et al., 2011; Pont et al., 2018).
eDNA analyses have been developed for about 10 years to survey freshwater biodiversity (eg. amongst others Ficeola, Miaud, Pompanon & Taberlet, 2008; Dejean et al., 2011; Taberlet et al. 2012; Darling & Mahon, 2012… see Thomsen & Willerslev, 2015 for a review). Most studies are still based on single-species detection and generally have a limited sampling plan (e.g., Goldberg, Sepulveda, Ray, Baumgardt, & Waits, 2013; Evans et al., 2015; Stoeckle, Huehn, & Geist, 2016; De Ventura, Kopp, Seppälä, & Jokela, 2017; Klymus, Marshall, & Stepien, 2017). Here, we present the first extensive dataset using standardized eDNA metabarcoding protocols for freshwater bivalves surveys, at the scale of whole France, with about 350 sampling sites. The present results reveal some inaccuracies in our previous knowledge of species distributions. As the extent and changes in distribution are one of the main parameters used for species conservation status assessments, this change of paradigm may have a strong impact on future conservation policies.